Sectoral Patterns Of Small Firm Innovation, Networking And Proximity

Research Policy 32 (2003) 751–770

Sectoral patterns of small firm innovation, networking and proximity Mark S. Freel∗ Department of Management Studies, Centre for Entrepreneurship, University of Aberdeen, Edward Wright Building, Old Aberdeen AB24 3QY, UK Received 28 September 2001; received in revised form 4 April 2002; accepted 5 June 2002

Abstract Drawing upon a sample of 597 small and medium-sized manufacturing firms, this article investigates the extent to which cooperation for innovation is associated with firm-level product and process ‘innovativeness’ and, where collaborative relationships are reported, the factors which influence their spatial distribution. With respect to the former issue, the data suggests considerable variety of association across Pavitt’s [Research Policy 13 (1994) 343] sectoral taxonomy and innovation type. However, the data also indicates the need for caution when developing network strategies or policies: the evidence presented here is unequivocal in noting that innovation is neither a necessary nor less a sufficient condition for innovation. Moreover, internal resources often act as complements to, or indeed appears to negate the need for, external resources. With regards to the spatial distribution of firm linkages, it appears that increasing firms size and export propensity are positively associated with external linkages at a higher spatial level. Moreover, the spatial reach of innovation-related linkages is also likely to be greater for firms reporting the introduction of relatively novel innovations (i.e. products or processes which are new to the industry). In contrast, smaller firms and firms engaged in incremental product innovations appear more likely to be locally embedded. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Innovation; Small firms; External linkages; Embeddedness

1. Introduction Following the work of Lundvall (1995), a consensus has emerged which holds that ‘. . . interactive learning and collective entrepreneurship are fundamental to the process of innovation’ (Lundvall, 1995, p. 9). This view marks a significant shift from the traditional view of innovation as a linear process spanning technology development activities and new product introduction (initiated by either scientific discovery (‘science-push innovation’) or market demand (‘demand-pull innovation’)). In this new conception, ∗ Tel.: +44-1224-274357; fax: +44-1224-273843. E-mail address: [email protected] (M.S. Freel).

innovation may no longer be understood as the outcome of independent decision-making at the level of the firm, but rather must be viewed as an iterative, cumulative and cooperative phenomenon, which incorporates more than simple phased dyadic or bilateral interactions between users, industry and the science base (such as those in envisaged by the classic Kline and Rosenberg (1986) chain-link model) (de la Mothe and Paquet, 1998). That is, since innovation involves the cumulative or path dependent creation of new knowledge, or novel recombination of existing knowledge, innovation is essentially concerned with learning. Learning, in its turn, is largely a social process—most especially in the context of the transfer and accumulation of tacit knowledge (Polanyi, 1966;

0048-7333/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 4 8 - 7 3 3 3 ( 0 2 ) 0 0 0 8 4 - 7

752

M.S. Freel / Research Policy 32 (2003) 751–770

Howells, 1995)—and is likely to involve considerably more than two actors. Accordingly, any innovation system is liable to be socially embedded in such a way that the innovativeness of individual firms will be influenced by socially specific, and extra-organisational, factors such as the infrastructure of financial institutions, labour markets, policy and provision of workforce training, mechanisms governing the support of business start-ups and development, attitudes and policy concerned with science and technology and the potency and pervasiveness of inter-firm and firm–institutional interaction (Dosi, 1999). It is this final issue, the role of external innovation-related cooperation, which this current paper seeks to address. Drawing upon data from a new survey of ‘Northern British’ SMEs, the paper is able to investigate the nature of associations between a variety of potential collaboration partners, the geographic location of these partners and the innovativeness of individual firms.

2. Innovation and networking Academics and policy makers rarely understate the importance of industrial innovation. In a world where such consensuses are rare, the common view holds that, ‘[i]n all highly-industrialised nations the long-term growth of business and (thus) of regions [and, one may safely assume, nations] stems from their ability to continually develop and produce innovative products’ (Sternberg, 2000, p. 391). Indeed, Chris Freeman, the doyen of innovation theorists, goes further; suggesting that ‘. . . not to innovate is to die’ (Freeman and Soete, 1997, p. 266). Whilst one may quibble about the appropriateness of such imperatives, it is nonetheless clear that innovations, of varying scale and scope, positively impact upon the performance of firms in aggregate and, by implication, economies (Geroski and Machin, 1992, 1993). Yet, the dominant network theory of innovation, in its many incarnations (cf. Håkansson, 1987; Maillat, 1995; Florida, 1995; Baptista and Swann, 1998; Cooke and Morgan, 1998; Oughton and Whittam, 1997), holds that individual firms are seldom capable of innovating independently, and never innovate in a vacuum. Frequently, advocates of network approaches to innovation induced economic development highlight an

increasing division of labour among organisations as a first principle compelling collaboration or interaction (Sternberg, 2000). That is, increasing uncertainty, associated with changing technology and global competition, has encouraged many firms (and, indeed, many nations) to concentrate on fewer and fewer core competencies, relying upon trade, or cooperation, for others (Archibugi et al., 1999). This effective disintegration of the vertical value chain has been taken as evidence of a move from hierarchical governance structures (based upon threat and coercion) to network governance structures (based up reciprocity and trust) (Nelson, 2000). In contrast to the pre-eminence of Fordist and Taylorist practices during the immediate post-war period, the efficient organisation of production is increasingly associated with vertical disintegration and flexibility (Hansen, 1990; Lawson, 1999). Fundamentally, however, the conclusion one is asked to reach, is that ‘. . . a high level of cooperation and communication typically affects innovation activities positively’ (Arndt and Sternberg, 2000, p. 467). In this regard, the instrumental benefits of inter-firm or inter-organisational collaboration are thought to revolve around the amelioration of internal resource constraints or competency gaps. In other words, a firm’s capacity for innovation will inevitably be enhanced by the extended knowledge base, and cost and risk sharing, offered through extensive linkages with external agency (e.g. suppliers, customers, competitors, universities, public agencies and so on) (see Rothwell (1991) and Freel (2000) for a more detailed discussion). This view, that no firm may function efficiently as an island entire of itself, has become somewhat axiomatic and has, in its turn, considerably influenced European industrial policy. For instance, amongst its principal conclusions, the most recent UK Competitiveness White Paper noted that: Firms may have to collaborate and network more, particularly on new technologies, and strengthen their links to the science and engineering base, sharing equipment and often people. (DTI, 1998, p. 29) Indeed, casual observation would suggest that UK competitiveness policy, throughout the 1990s, has been premised, to a greater extent, upon the belief that there exist inadequate linkages between firms, and between industry and the science base (Oughton, 1997).

M.S. Freel / Research Policy 32 (2003) 751–770

Regardless of the sway held, recent empirical work by Oerlemans et al. (1998, p. 300) suggests an important caveat to the foregoing; that the network view of innovation ‘. . . overemphasises an inter-organisational approach to organisational processes’. In our haste to embrace the importance of external linkages in facilitating firm-level innovation, there may be a tendency to neglect the contribution made by internal resources. Yet, in most industries, the greater part of innovation effort is made by firms themselves (Nelson, 2000) and occurs within firms themselves. As Dosi notes, quite emphatically, firms require ‘. . . substantial in-house capacity in order to recognize, evaluate, negotiate, and finally adapt the technology potentially available from others’ (Dosi, 1988, p. 1132; see also, Cohen and Levinthal, 1989, 1990). In other words, the interactivity of the innovation process refers to collaborations and iterations involving individuals and departments within the firms as well as to, more occasionally, external co-operations with other organisations and institutions (Freel, 2002). This former point is often neglected. Though this is not to suggest the Håkanssonian (1987) view that ‘. . . the company is interlocked with others—it is not a free and independent unit’, is in any sense erroneous. Merely, that the notion of ‘interlocking’ does not necessarily imply cooperation, in any robust sense, between companies.

3. Proximity and networking In terms of the role of spatial proximity in facilitating inter-firm (and firm–institution) collaboration and information exchange, much of the impetus has been provided by the conceptual developments stemming from the ‘new industrial districts’ literature and the success enjoyed by empirical exemplars such as the ‘Third Italy’ (see for example, Brusco, 1982; Becattini, 1978, 1990; Bianchi, 1990). This literature defines innovation networks, to a greater extent, spatially—viz. ‘. . . a territorial agglomeration of small firms, normally specialised by product type, product components or process phases, held together by inter-personal links, by common “social culture” amongst the workers, entrepreneurs and politicians and enveloped by an “industrial atmosphere” . . . ’ (Bianchi, 1998, p. 96). Whilst not explicitly precluding

753

extra-regional links, one is, nonetheless, given to understand that the ‘district’ has geographic boundaries. The models of regional economic development, which have arisen subsequently: . . . emphasise the spatial organisation of the market by different players (firms essentially), the inter-relation between these players and, eventually, the diffusion of economic growth from a given set of players to the rest of the geographic area. (Andréosso-O’Callaghan, 2000, pp. 70–71) Even in their larger conception (i.e. incorporating interactions between cultural, social, political and economic actors), these conventional models of organisational and economic development require implicit spatial/geographic boundaries and are, in this sense, closed systems. The central issue, as Howells (1999) records, is that: ‘Studies indicate to a lesser or greater extent a typical distance decay function in communication’ (p. 83). Yet, proximity is not simply a spatial phenomenon. Whilst, historically, physical barriers (such as oceans, mountains or deserts) made it difficult for individuals to learn about opportunities or share in first mover advantages, one may reasonably anticipate that recent developments in ICT will have significantly reduced this ‘tyranny of distance’. Indeed, in the modern global economy, politically inspired borders between countries and regions may create regulatory hurdles that serve to constrain the flow of goods, services and, crucially, information (Brown and Butler, 1993). As Staber (1997, p. 59) notes, whilst ‘[g]eographic proximity is important to the extent that it facilitates information exchange and mutual adjustment’, information is not the only commodity and proximity may also be defined socially—and, indeed, in terms of technology, organisation or institutional framework (see Kirat and Lung, 1999). Thus, the relevant organisational population may, more appropriately be defined as ‘. . . a set of organisations similar in their structures and processes and in their relations with actors in their environment, but with no theoretical limit placed on the territorial location of populations’ (Staber, 1997, p. 59; tense changed). This is not to suggest that spatial proximity is of no importance in influencing the level of interaction between economic actors. Indeed, with respect to university linkages, Freel (2000) notes the importance

754

M.S. Freel / Research Policy 32 (2003) 751–770

of personal contacts, with adjacent firms, in the diffusion of knowledge (see also Acs et al., 1994). However, it is likely that there is an inverse relationship between cognitive proximity (and, relatedly, the extent to which the knowledge involved is tacit or codifiable) and the importance of spatial proximity. Nooteboom’s (1999) notion of ‘cognitive proximity’ draws on both Granovetter’s (1982) ‘strength of weak ties’ and Cohen and Levinthal’s (1990) ‘absorptive capacity’. In this way ‘cognitively distant’ knowledge has ‘the advantage of yielding more novelty or ‘non-redundance’, but understanding may be complicated by a lack of shared experience. In particular, when knowledge is tacit, strong ties, based on enduring and intensive interaction, may be needed’ (Nooteboom, 1999, p. 140). Simplistically, following Cohen and Levinthal (1990), knowledge is considered cumulative and more difficult in novel domains. In this sense we are viewing the transfer of knowledge between organisations as a process of ‘learning by interacting’ (Lundvall, 1995)—as distinct from the ‘learning by doing’ (Arrow, 1962) and ‘learning by using’ (Rosenberg, 1982). Accordingly, ‘. . . the more complex the learning process, the more interactions it probably requires’ (Johnson and Lundvall, 1993, p. 75) and, presumably, the more necessary spatial proximity becomes—in facilitating frequency of direct interaction. To restate, if the requisite knowledge is cognitively distant from the firm’s internal knowledge base (which, in turn, determines its ‘absorptive capacity’) then spatial proximity becomes important in assisting effective knowledge transfer. The reverse is also likely to hold and, while the potential impact of spatial proximity is unlikely to tend to zero, in the presence of cognitively proximate (in the sense of shared frames of reference, experiences and mental models), highly-targeted (to the needs and concerns of the firm) and codifiable information, its impact upon the innovation process diminishes. However, it is likely that, for the majority or firms, the knowledge that is required for innovation will be relatively cognitively proximate. Such firms invariably occupy mature industries and are, at best, characterised by a flow of incremental, rather than radical, innovations (Karlsson and Olsson, 1998)—which build immediately upon their current state-of-the-art. As Lundvall insists, ‘[t]he first step in recognizing innovation as a ubiquitous phenomenon is to focus

upon its gradual and cumulative aspects . . . Almost all innovations reflect already existing knowledge, combined in new ways’ (Lundvall, 1995, p. 8). Accordingly, much of this knowledge is likely to be available, and much of the learning will take place, internally. Where internal knowledge deficiencies are identified and external know-how is required, one may reasonable propose that, more often than not, organisational and institutional proximities will play a more significant role, than spatial proximity, in defining the nature of external articulation. Organisational proximity, for instance, relies upon commonalities in ‘. . . the representations and structures which agents use as a benchmark in order to define both their routines and strategic practices’ (Kirat and Lung, 1999, p. 30). To the extent that organisational proximity arises ‘. . . between organisations connected by a relationship of either economic or financial dependence/interdependence’ (Kirat and Lung, 1999, p. 30), intra-industry links are liable to dominate inter-industry links. Institutional proximity, by contrast, implies a degree of congruence between, and acceptance of the legitimacy of, the institutional infrastructure in which agents operate. And, in turn, the impact the institutional framework has upon the development of cognitive models. If the innovation process requires ‘. . . the development of cognitive models which are both coherent and shared’ (Kirat and Lung, 1999, p. 28) (in terms of structure rather than content) we might anticipate firms embedded within similar institutional frameworks (political, financial, economic, cultural) to be more readily disposed to collaboration. In addition to the influence of various proximities, recent evidence has pointed to the influence of firm size and product market on the spatial distribution of innovation-related external linkages (Arndt and Sternberg, 2000; Kaufmann and Todtling, 2000). Specifically, these studies suggest a positive relationship between firm size and export intensity and the geographic reach of search and collaboration activities. In other words, it appears that, in aggregate, smaller firms are more spatially embedded than larger firms and non-exporters are more embedded than exporters. Thus, integration within local networks may not be the outcome of conscious decision-making, on the part of the firm (or, relatedly, any attempt at some form of optimisation), but rather may indicate either resource limitations, which constrain the reach of firm

M.S. Freel / Research Policy 32 (2003) 751–770

search processes, or development stage, which sees the firm serve a largely, and initially, localised market. In either instance, the features of a firm’s innovation process or the pervasiveness of some ‘local’ institutional innovation system may have limited influence on the spatial composition of its external linkages. In light of these emerging tensions in the literature, the object here is to address a number of related empirical questions: Question 1a: To what extent is the innovation performance of firms associated with articulation (beyond simple transaction relationships) with external agency? The following question, offering a variety of potential collaboration partners, was used to determine the presence of ‘linkages’: ‘Did your firm co-operate with other firms or organisations for innovation related activity (including marketing, training, etc.) and/or technology transfer during the last 3 years?’ Question 1b: What is the relative importance of various potential collaboration partners? Question 1c: What is the relative importance of external linkages and internal resources? Question 1d: How do these issues differ across sectoral groupings? Question 2: Where innovation-related external linkages are reported, to what extent are they clustered locally and what factors determine their spatial distribution? 4. Data 4.1. Sample The data presented here was collected as part of a ‘Survey of Enterprise in Scotland and Northern England’. This original survey, administered during April and May of 2001, sought to gain a better understanding of the state of (small and medium-sized) enterprise within the UK, sufficiently removed from the thrall of the dominant, core economy of London and the Southeast. To this end, the original sample frame included both manufacturing and service firms. However, that the nature of innovation (at the heart of the current paper) varies substantially and systematically between manufacturing and service firms is fairly well established (see for example, Hoffman

755

et al., 1998).1 Accordingly, the current analysis sets aside service firms and focuses upon manufacturing firms in isolation. Specifically, 5200 small and medium-sized (FTEs <500) manufacturing firms were surveyed, generating 597 usable responses (i.e. a 11.5% response rate). As Fig. 1 illustrates, the sample is stratified in such a way as to under-represent the smallest firms (one to nine employees) and over-represent larger SMEs. Surveys of micro-firms tend to report lower response rates and, in the case of innovation related surveys, poorer data quality (see Cosh et al., 1998). Undoubtedly, this deliberate skew impacts upon the ‘representativeness’ of the current sample. The chief consequence is likely to be an over-estimate of the level of innovation within the population. However, it is not anticipated that this ‘bias’ will greatly impact upon the analysis presented below. With regards to the relative sectoral distribution of sample firms, Fig. 2 indicates an over-surveying of textile and clothing firms and firms involved in the manufacture of wood and paper products (SIC (92) Divisions 17–19 and 20–21, respectively)—relative to the pertinent population. By contrast, printing and publishing firms and manufacturing firms not elsewhere specified (NES) are significantly under-represented in the sample (SIC (92) Divisions 22 and 36–37, respectively). In all other SIC (92) divisions, the sample reasonably represents the population—though, in the case of food and beverages, not the UK. The decision to over-survey firms in the identified sectors was made with reference to recognised regional industrial clusters (DTI, 2001a). Over-representation of some sectors necessarily leads to under-representation of others. Accordingly, the decision was made to under-survey printing and publishing firms—since many of these are de facto business service providers—and manufacturing firms ‘not elsewhere specified’—as a result of the ambiguity inherent at the two-digit level. Whilst stratification of the sample, in this manner, will inevitably distort aggregate observations, one does not anticipate that the legitimacy of the analysis presented in Section 5 will be greatly compromised. 1

Although this is not to imply that service firms merely play a passive role [as users] in the process of technological development. Simply, that to lump service firms and manufacturing firms together, for the purposes of analysis, is to ignore the various dynamics in play.

756

M.S. Freel / Research Policy 32 (2003) 751–770

Fig. 1. Relative size distribution of sample firms.

5. Empirical analysis 5.1. Modelling innovation In investigating the extent to which external linkages (as defined earlier) are positively associated with firm-level innovative performance, the current paper follows the established practice (see Geroski, 1990; Feldman, 1994; Love and Roper, 1999, 2001; Oerlemans et al., 2001) of modelling innovation output using a modified knowledge production function approach. In this way ‘. . . innovation output depends on the presence and volume of innovation resources and the utilisation of these internal and external resources in the innovation process’ (Oerlemans et al., 2001, p. 9): INNi = β1 RDi + β2 TECHi + β3 QSEi +β4 CUSTi + β5 SUPPi + β6 COMPi +β7 UNIi + β8 PUBi + ε In addition, the analysis reported later on incorporates two moderating variables: age (measured in years) and firm size (full-time equivalent employees). On the whole, one may reasonably suggest that younger firms are less likely to have established routines, technology or products and are less likely to occupy mature or traditional industries. Moreover, with regards firm size, whilst the evidence is decidedly

equivocal, recent academic opinion runs counter to the established Schumpeterian or Galbraithian view (that increased market concentration and, relatedly, increased firm size act as necessary stimuli to innovation) and holds instead that, subject to sectoral variations, small firms may, in fact, be more innovation intensive than large firms (Acs and Audretsch, 1987, 1988). Therefore, adopting a very general view of innovation, one might anticipate that younger and smaller firms report proportionately higher levels of innovation than their older and larger peers. As Table 1 indicates, innovation has been defined using an absolute and output-based measure. In the estimations firms are classed as innovators only if they record the introduction of, at least one, new product or process, which is new, not only to the firm but also, to the industry. In this way, minor product or process improvements or mere competitor imitation are excluded. However, the degree of novelty implied by this measure is relative. One would anticipate that, in the bulk of mature industries, such product and process innovations would, nevertheless, build closely upon the firm’s current technological state-of-the-art. This position is also in line with the view of innovation as a commercial endeavour rather than a purely technological endeavour and accords with the observation that bulk of commercially significant innovations in manufacturing are technologically incremental rather than radical (Audretsch, 1995; Love and Roper, 1999).

M.S. Freel / Research Policy 32 (2003) 751–770

Fig. 2. Relative sectoral distribution of sample firms.

757

758

M.S. Freel / Research Policy 32 (2003) 751–770

Table 1 Variables used in logit equations Definition of Variables INN RD TECH QSE CUST SUPP COMP UNI PUB

Binary dummies of product and process innovation; firm introduced at least one product or process new to the industry ‘during the last 3 years’ coded 1, otherwise, coded 0 A series of binary dummies of research and development expenditure as a proportion of sales turnover; categories are 1–5%, 6–10%, 11–20% and >20%; no R&D spend is treated as the reference group Proportion of workforce (FTEs) classed a technicians Proportion of workforce (FTEs) classed as technologists or scientists Binary dummy of customer focused innovation networking; if firm cooperated with a customer for innovation-related activities ‘during the last 3 years’ coded 1, otherwise, coded 0 Binary dummy of supplier focused innovation networking; if firm cooperated with a supplier for innovation-related activities ‘during the last 3 years’ coded 1, otherwise, coded 0 Binary dummy of competitor focused innovation networking; if firm cooperated with a competitor for innovation-related activities ‘during the last 3 years’ coded 1, otherwise, coded 0 Binary dummy of university focused innovation networking; if firm cooperated with a university for innovation-related activities ‘during the last 3 years’ coded 1, otherwise, coded 0 Binary dummy of public sector focused innovation networking; if firm cooperated with a public agency (e.g. UK government offices, EU, enterprise companies/agencies, etc.) for innovation-related activities ‘during the last 3 years’ coded 1, otherwise, coded 0

Finally, sectoral variations relating to innovation are fairly well established. For instance, it is known that industrial sectors vary in terms of the sources, paces and rates of technological change (Pavitt, 1984). Moreover, ‘[b]ecause sectoral patterns of technological innovation are different, one may expect that firms in specific sectors use specific internal and external resources in order to innovate successfully’ (Oerlemans et al., 1998, p. 302). That is, sectoral variations are likely, in their turn, to modulate the requirement for firms to be engaged in innovation networks and, relatedly, the extent and character of such networking. In attempting to control for ‘sectoral noise’, the current paper adopts Pavitt’s (1984) classic taxonomy, ‘supplier-dominated’ firms, ‘production-intensive’ firms, and ‘science-based’ firms. In his paper, Pavitt further disaggregates production-intensive firms into ‘scale-intensive’ and ‘specialist suppliers’. However, since the current study is exclusively concerned with SMEs (less than 250 full-time employees), it is likely that the majority of firms classed (at the two-digit SIC (92) level) as scale-intensive will be de facto suppliers of components, sub-assemblies and equipment to genuinely scale-intensive firms. Consequently, the models estimated in Section 5.2, recombine these sub-categories for the purposes of analysis.

5.2. Anticipated sectoral variations Prior to detailing the results of the analysis, it seems prudent to outline a priori anticipations of sectoral variations in external and internal resources usage. 5.2.1. Supplier-dominated firms The innovation focus of such firms is thought, largely, to be concentrated on cost reducing process technologies to meet the demands of highly price-sensitive customers. However, given the generally weak in-house R&D and engineering capabilities, suppliers are the likely source of new or improved process technologies. Notwithstanding the relative emphasis on process innovation, Pavitt (1984) suggests that general sources of technology (including those which lead to product innovations) are liable to include, in addition to suppliers, government financed research extension services and, less frequently, large users. On the whole, however, given the greater propensity to sell to myriad end users, or through distribution agents, one would not anticipate a significant role for customers. Competitor collaboration, of the ‘Third Italy’ type, may be present; though this is hope more than anticipation. Finally, since supplier-dominated firms are believed to ‘make only a minor contribution to their process and product

M.S. Freel / Research Policy 32 (2003) 751–770

technology’ (Pavitt, 1984, p. 356) one would anticipate a limited association between internal resources and innovation.

759

technology is largely developed in-house or sourced from suppliers, whilst product technology is extended internally ‘. . . based upon the rapid development of the underlying sciences in the universities and elsewhere’ (Pavitt, 1984, p. 362). This suggests a role for universities in product innovation and suppliers in process innovation, which complement extensive in-house capabilities. Innovation-related collaboration with customers, competitors and (quasi-) government agencies is likely to be limited.

5.2.2. Production-intensive firms Given the earlier contention that, in the current sample, these firms are largely de facto suppliers to genuinely scale-intensive firms, and the relative focus on product innovation, one would anticipate a positive association between innovativeness and customer collaboration—at least in the case of product innovation. In contrast, suppliers and the public knowledge infrastructure are liable to play a limited role. Some competitor collaboration may be present as a result of supply chain dicta or management. Finally, in line with Pavitt’s observations, one might expect in-house capabilities to make a significant contribution to both product and process innovations.

5.3. Analysis Estimation of the models, suggested in Section 5.2, takes the form of eight logit equations, employing ‘new [to the industry] product introduction’ (Table 2) and ‘new [to the industry] process introduction’ (Table 3) as binary dependent variables. On the whole, the models seem to have a number of satisfactory properties. For instance, various tests for multi-collinearity (using correlation matrices, and multiway frequency analysis (Tabachnik and Fidell, 2001)) suggest

5.2.3. Science-based firms The innovation focus of many science-based firms is a balance between products and processes. Process

Table 2 Logit models of the probability of introducing new (to industry) products Independent variables

Pavitt (1984) sectors Total sample

Age Size (FTEs) R&D expenditure 1–5% of turnover 6–10% of turnover 11–20% of turnover >20% of turnover Percentage of technicians Percentage of QSEs Customer links Supplier links Competitor links University links Public sector links

0.707 0.622 1.272 0.683 1.701 1.565 0.565 −0.046 −0.053 0.230 0.534

Nagelkerke R2 −2 log-likelihood (χ2 )d N

0.192 629.874 80.664a 528

a

0.00 (0.011) 0.008 (19.913)a (10.618)a (2.867)c (6.317)b (1.707) (4.292)b (2.122) (5.735)b (0.036) (0.031) (0.416) (3.846)b

Supplier-dominated

Production-intensive

0.00 (0.004) 0.014 (5.362)b 1.705 1.954 2.918 2.107 2.490 −6.015 −0.075 1.322 1.088 −9.808 −0.455

(6.357)b (3.984)b (4.552)b (2.824)c (2.103) (0.107) (0.013) (3.990)b (1.616) (0.071) (0.368)

0.400 98.064 34.468a 98

Significant at 1% level. Significant at 5% level. c Significant at 10% level. d Full model vs. constant only model. Figures in parentheses are Wald χ2 -test statistics. b

−0.001 (0.042) 0.009 (14.335)a 0.638 0.199 0.578 0.298 1.300 7.099 0.869 −0.252 −0.147 0.010 0.910 0.230 390.275 62.679a 337

(5.618)b (0.137) (0.770) (0.149) (1.233) (3.261)c (8.590)a (0.674) (0.153) (0.001) (7.362)a

Science-based −0.002 (0.070) 0.008 (2.658)c 0.017 1.009 1.597 0.745 1.612 2.097 0.177 −0.217 −0.357 2.065 −1.185 0.227 107.921 17.095 93

(0.001) (1.621) (1.377) (0.293) (0.542) (1.585) (0.027) (0.092) (0.195) (3.202)c (1.401)

760

M.S. Freel / Research Policy 32 (2003) 751–770

Table 3 Logit models of the probability of introducing new (to industry) processes Independent variables

Pavitt (1984) sectors Total sample

Age Size (FTEs) R&D expenditure 1–5% of turnover 6–10% of turnover 11–20% of turnover >20% of turnover Percentage of technicians Percentage of QSEs Customer links Supplier links Competitor links University links Public sector links

1.031 0.042 1.359 0.952 2.275 −1.430 −0.087 0.689 −0.320 0.929 0.051

Nagelkerke R2 −2 log-likelihood (χ2 )d N

0.142 418.321 45.571a 522

Supplier-dominated −0.002 (0.018) 0.009 (2.585)c

0.001 (0.023) 0.004 (4.852)b (11.337)a (0.006) (5.912)b (2.183) (6.134)b (0.474) (0.079) (5.220)b (0.758) (5.606)b (0.022)

Production-intensive

0.922 −1.206 1.773 11.837 0.133 −111.890 −22.020 1.968 0.864 21.309 −10.942

(0.974) (0.203) (1.352) (0.028) (0.002) (0.267) (0.052) (3.084)c (0.262) (0.002) (0.016)

0.498 43.050 29.555a 97

−0.001 (0.047) 0.003 (1.720) 1.501 0.158 1.489 1.266 3.791 −0.072 0.502 0.559 −0.437 0.540 0.287 0.210 260.713 44.889a 335

(13.754)a (0.034) (3.895)b (1.824) (8.862)a (0.000) (1.804) (2.255) (0.955) (1.458) (0.496)

Science-based 0.008 (0.724) 0.009 (3.162)c −0.201 −0.746 0.995 −6.214 −1.421 −2.897 −1.336 0.730 −0.730 2.500 −0.897

(0.065) (0.374) (0.485) (0.033) (0.139) (0.638) (1.361) (0.666) (0.406) (2.919)c (0.466)

0.263 68.585 15.656 90

a

Significant at 1% level. Significant at 5% level. c Significant at 10% level. d Full model vs. constant only model. Figures in parentheses are Wald χ2 -test statistics. b

little problem in this respect. Moreover, as the data in Tables 2 and 3 indicates, the models appear reasonable predictors of ‘innovativeness’ for all but science-based firms (see later). With reference to the moderating variables identified before; innovators seem, on the whole, to be larger than non-innovators, which runs counter to the stated expectations. However, given the definition of innovation employed and the sectors that dominate the sample, this is perhaps less remarkable than if one were investigating innovation intensity in less mature, embryonic industries (Acs and Audretsch, 1987). In contrast, firm age appears to have no impact upon the ‘innovativeness’ of firms. Addressing product innovation, in the first instance, and turning to specifics, the all-sample model indicates significant positive associations between the firm’s innovation-related cooperation with customers and public (and quasi-public) sector agencies and innovative performance. Collaboration with customers for product innovation underpins von Hippel’s (1978) ‘Customer-Active’ innovation paradigm (CAP), wherein ‘continuous user–manufacturer interaction

. . . identifies re-innovation opportunities, new uses and new users’ (Shaw, 1991, p. 127; tense changed). Moreover, it is commonly accepted that there is scope for considerable gain through involving the user in the product design and development processes. In the context of the current study, such gains are believed to be principally three-fold (see Gardiner and Rothwell, 1985): firstly, firms may be able to supplement their internal design and development activities by accessing the technical and managerial skills resident within their customers; secondly, user involvement is likely to be the ideal way to establish the optimum price/performance combination and, consequently, the optimum specification; finally, involving the user in the product design and development stages is likely to reduce the post-delivery learning required on their part (and accordingly, this may result in strong demonstration effects, attracting other customers and accelerating the innovation acceptance process). With regards to public sector linkages, the data may provide some legitimating evidence with regards to the role of public agencies as network facilitators: in

M.S. Freel / Research Policy 32 (2003) 751–770

the ‘. . . encouragement of networking relationships between firms to establish collective economies of scale’ (Oughton and Whittam, 1997). Regardless, the all-sample observations paint a superficially encouraging picture of the potential contribution that collaboration with public agency may play in facilitating innovation. However, it is important to note the significant role played by internal resources: product innovation appears to be positively associated with the employment of technicians and with research and development expenditure—up to a point. This issue of ‘up to a point’ is interesting, perhaps suggesting some form of decreasing returns to R&D investment at the firm level. However, this would be to endorse the conventional economics view of R&D: namely, that it generates only one output—new information or knowledge (frequently embodied in physical artefacts, such as new or improved products). Following Cohen and Levinthal (1989, 1990) this is generally believed to be an erroneous position, though the finding is, in itself, intriguing. Finally, indeed crucially, notwithstanding the positive observations noted for customers and the public sector, innovation-related cooperation with suppliers, competitors or universities does not appear to be linked to the product innovativeness of sample firms in aggregate. With regards to expected sectoral variations: the positive relationship with external cooperation (involving customers and public sector agencies) does not appear to hold across Pavitt’s taxonomy. For instance, in the case of ‘supplier-dominated’ firms, and largely in line with expectations, only supplier collaboration is positively associated with product innovation. Moreover, this appears to be complemented by in-house R&D expenditure (of any magnitude). Conversely, there is no statistical relationship between the product innovativeness of supplier-dominated firms and collaboration with customers, universities, competitors or the public sector. Perhaps unsurprisingly, given their predominance in the current sample, the pattern of observations for production-intensive firms is similar to that noted for the all-sample estimations. In particular, with respect to innovation-related cooperation, one notes positive associations between product innovativeness and collaborative linkages with customers and with the public sector. The former observation is in line with stated

761

expectations: accepting the presumed focus upon performance improving product innovation, rather than cost reducing process innovation, there is likely to be scope for customers to play a key role in product design and specification activities. The latter observation (i.e. a strong positive association between product innovativeness and cooperation with the public sector) is less anticipated. However, it may reflect a determination, on the part of government, to develop the innovativeness of lower tier suppliers as a means to strengthening supply chains and improving competitiveness (DTI, 2001b). Significantly, these external linkages appear to be complemented by internal resources, in the form of (limited) R&D expenditure and employment of QSEs. Finally, as with the all-sample model, there are no statistical relationships between firm-level innovativeness and cooperative ventures involving competitors, universities or suppliers. With respect to the final sectoral grouping, science-based firms, the data in Table 2 indicate only a weak significant association between product innovation and university collaboration: firm size apart, no other significant relationships are observed. That product innovation in science-based firms is positively associated with university cooperative ventures is as anticipated. Indeed, the absence of observed statistical associations between innovativeness and customers, competitor, supplier or public sector collaboration may also be in line with expectations. However, lack of observed statistical relationships with R&D expenditure or the employment of QSEs is more remarkable and runs counter to both intuition and empirical precedent. Moreover, in this instance, the model itself does not appear to be an effective predictor of product innovativeness (in the absence of statistical significance). This observation may suggest some source of firm heterogeneity beyond that captured by the regressors. It may be that the manner in which resources are developed, extended or combined (i.e. transformation activities, see Håkansson (1987)) is of greater import than simply the volume and type of resource input. Unfortunately, measurement of such transformation activities is beyond the scope of the current study. Turning, next, to process innovation (Table 3): Perhaps unsurprisingly, the observed statistical associations between the introduction of a (new to the industry) process innovation and external collaboration

762

M.S. Freel / Research Policy 32 (2003) 751–770

varies markedly from those recorded for product innovation. For instance, and with regards to the all-sample model in the first instance, innovation-related cooperation with customers or public sector agencies is no longer significantly associated with innovativeness. Conversely, one observes statistically significant relationships relating to supplier and university collaboration. Undoubtedly, there is a comfortable logic in this reversal. To suggest that suppliers and customers have greater roles to play in process and product innovation respectively is likely to be uncontroversial and, in this sense, is reassuring. The association involving university collaboration and firm-level process innovation is also encouraging and may indicate the greater role that universities play in addressing the immediate concerns of industry than is often thought. As Nelson (2000) asserts, ‘[t]o a much greater extent than commonly realized, university research programmes are not undifferentiated parts of a national innovation system broadly defined, but rather are keyed into particular technologies and particular industries’ (p. 13). Crucially, one also notes the complementary correlations between process innovativeness and internal resources; proxied by firm size, R&D expenditure (to a point) and the employment of technically qualified staff. Notwithstanding this general picture, the data also indicate considerable sectoral variations. In line with expectations, successful new process introduction by supplier-dominated firms is strongly associated with supplier cooperation. Given presumptions of poorer in-house R &D and engineering facility, in such firms, the absence of significant observations linking internal resources and process innovation is also as anticipated. Speculating from the discussion of product innovation given earlier, it may be that where R&D is present within supplier-dominated firms, it is directed towards product innovation, with significant process innovation largely driven by equipment suppliers. For the bulk of the current sample, the productionintensive firms, Table 3 indicates no statistical relationships between successful new process introduction and external innovation-related collaboration. Again, this is largely in line with expectations. Whilst the analogous model for product innovation suggested a positive statistical association involving customer (and public sector) collaboration, these firms are thought to ‘. . . produce a relatively high proportion of

their own process technology’ (Pavitt, 1984, p. 359). Accordingly, the significant positive observations involving R&D expenditure and technical employment are as anticipated: in this instance, in-house resources appear to play a greater role. Finally, with respect to science-based firms, similar observations are noted for both product and process innovation. That is, significant correlations between innovativeness and university collaboration and innovativeness and firm size only (both significant at the margin). Furthermore, the model is, again, a poor predictor of new process introduction. Consequently, one is, once again, tempted to suggest the existence of some form of firm heterogeneity not captured by the regressors: perhaps signalling the importance of varied transformation activities, rather than the mere presence of heterogeneous resources. On the whole, the empirical data indicates a number of relevant considerations. In the first instance, one might be tempted to suggest that, in general, the data denotes a positive association between innovation-related cooperation and firm-level innovativeness—i.e. networks matter. However, this is misleading and any statement must, more appropriately, be couched in conditional terms. That is, certain types of cooperation are associated with specific types of innovation, involving certain firms, in certain sectors. Conversely, many sources of innovation-related collaboration are not correlated with specific types of firm-level innovativeness, in certain sectors—indeed, competitor cooperation is not associated with either product or process innovativeness in any sectoral category. Clearly, blanket collaboration imperatives overstate the case and misunderstand the various dynamics at work. Moreover, the models presented indicate the frequent importance of internal resources, as either a complement to, or a substitute for, external collaboration. As noted earlier, in most industries the greater part of innovation effort is made by firms themselves (Nelson, 2000) and occurs within firms themselves. Indeed, even in circumstances where the requirement for collaborative effort is identified, it is thought essential that firms have developed internal competencies which facilitate the effective recognition, appraisal, negotiation and assimilation of external expertise (Dosi, 1988). Only rarely will firms be mere passive receptors of exogenous technology (broadly defined).

M.S. Freel / Research Policy 32 (2003) 751–770

Finally, there is some indication (at least in the case of science-based firms) that the standard knowledge production function approach may not accurately represent firm-level innovation—or proxy firm-level innovation processes. In other words, innovation outputs are not merely the sum of innovation inputs but, rather, are likely to also be a function of the activities through which the resource inputs are transformed. Standard approaches, such as that reported here, may not adequately capture these underlying dynamics. In addition to the conditions noted earlier, a further, central, caveat must be added. As Fig. 3 clearly demonstrates, a great many firms were engaged in no innovation-related external collaboration during the period investigated by the survey. And yet many of them appear to have successfully innovated nonetheless (e.g. 31.3% of firms with no external links recorded the introduction of a product new to the industry; though this is compared with 50.4% of firms with at least one innovation-related link). To rephrase, networking (in the sense of cooperation adopted here) cannot be considered either a necessary, nor less a sufficient condition for innovation and it may be that, within both the academy and the polity, the tendency to overstate the impact of networks persists. Furthermore, the vast majority of

763

all cooperative ventures reported occurred along the value chain. This accords with our earlier discussion regarding the nature of technological, organisational and institutional proximities. That is, for most firms, occupying relatively mature industries and engaged in technologically incremental innovations, the information shared, or exchanged, in relationships with customers and suppliers is likely to be relatively cognitively proximate and be facilitated by organisational (in the sense of similarities in routines, strategies and governance) and, within the UK (and perhaps Europe), institutional (in the sense of similarities in institutional infrastructure) proximities. This appears to hold equally for science-based firms. In other words, irrespective of sectoral classification, and the implications for innovation this implies, relatively few horizontal, competitor links, or extra-industry links were reported. This is in contrast with the simultaneous competition–collaboration expectations of much of the ‘new industrial districts’ literature (see for example, Oughton and Whittam, 1997) and the science-industry interactivity aspirations of public policy (Etzkowitz and Leydesdorff, 2000) and may suggest that the frequency of contact implied by market relations may prove a more robust basis for the development of trust and cooperation under uncertainty.

Fig. 3. Proportion of firms with innovation-related external linkages.

764

M.S. Freel / Research Policy 32 (2003) 751–770

5.4. Innovation and embeddedness The second question identified in the early parts of the paper concerned the spatial distribution of external, innovation-related linkages. Specifically, where linkages were reported, to what extent were these clustered spatially? To this end, Table 4 reports the results of a multiple-discriminant analysis2 with groupings based on the highest spatial level of innovation-related links (i.e. ‘local’, ‘regional’, ‘UK’ and ‘overseas’). Interpretation of the functions relies largely upon consideration of the correlations between individual predictors variables and the respective functions (in effect, factor loadings)— supported by the results of univariate ANOVAs for each variable. Whilst consensus is lacking with regards to the interpretation of loadings, a common ‘rule of thumb’ suggests that correlations in excess of 0.33 (i.e. 10% of variance) may be considered eligible for interpretation whilst lower ones may not (Tabachnik and Fidell, 2001). Accordingly, the following discussion adopts this convention. As the data indicates, the three functions generated are effective predictors of group membership (P = 0.002; see also Fig. 4). In particular, larger firms and firms with a higher export intensity are significantly more likely to have overseas links, whilst smaller firms are more likely to be locally embedded (see Table 4). As noted earlier, this is in accordance with other recent studies demonstrating the importance of firm resources, development stage and market orientation in determining the spatial distribution of innovation-related collaboration. The importance of internal resources is further underscored by the significant negative impact noted for the proportionate employment of managers and professionals. That is, a relatively higher proportion of staff designated as either managers or professionals appears to be associated with a higher degree of local embeddedness. This finding is somewhat counter-intuitive—one would expect the presence of significant managerial resource to positively impact upon the firm’s ability to interact with geographically distant organisations. However, treating higher levels of management as a proxy for 2

Initial estimations incorporated sector dummy variables relating to Pavitt’s taxonomy. However, these did not prove significant, and, indeed reduced the discriminating power of the functions.

increasing bureaucracy, one may propose that relatively managerialist firms are likely to suffer from less flexibility and be less attractive, as innovation partners, to firms operating in national or international markets. Alternatively, greater managerial resource may suggest a concomitant ability to more fully exploit local knowledge. However, in the absence of richer, ‘firm-specific’ data, one may merely conjecture upon the nature of causality—though, again, the finding is intriguing. In contrast to the negative effects of managerialism, though not supported by the univariate ANOVA, the multivariate data from function 1 indicates a significant positive association between the proportion of technicians in the workforce and the presence of extended spatial linkages. In other words, firms that recorded, at least one, innovation-related overseas cooperative venture, are also likely to record a higher proportion of technicians in the workforce than are firms with national or regional linkages only who, in turn, record higher proportions of technicians than locally embedded firms). This finding may simply reflect the converse of the negative impact (on the spatial reach of cooperation) observed in the presence of high levels of management. However, a high proportion of technically qualified staff may also hint at a more developed absorptive capacity, which extends the reach (both in distance and scope) of the firm’s cooperative potential. However, again in the absence of richer firm-specific data, this remains conjecture. With regards to the effect that type and scope of innovation have on the spatial distribution of external articulation, the data from the current study suggests a significant positive relationship between novelty and reach. That is, the proportion of firms engaged in relatively novel innovation (introducing products or processes which are new to the industry)3 increases as one progresses through the spatial groupings. By contrast, the proportion of firms engaged in incremental product innovation only (introducing products new to the firm but not to the industry) appears inversely related to the spatial categories contrived. Incremental process innovation appears to have no impact upon the spatial distribution of innovation-related linkages. To rephrase, novel innovators (product and process) 3 Type of innovation is coded using four binary dummy variables; firms cannot be classified as both novel and incremental innovators for either products or processes—in both instances ‘no innovation’ is the excluded category.

M.S. Freel / Research Policy 32 (2003) 751–770

765

Table 4 Discriminant functions of the spatial distribution of cooperative ventures Predictor variables

Correlations of predictor variables with discriminant functions 1

2

3

Univariate F(3, 248)

Firm age Size (FTEs) R&D expenditure Percentage of technicians Percentage of QSEs Percentage of professionals/managers Export (as % of T/O) Novel products Incremental products Novel processes Incremental processes Canonical R Eigen value

0.010 0.437 0.282 0.330 0.232 −0.445 0.573 0.488 −0.368 0.070 0.083 0.355 0.145

0.406 0.023 0.118 0.003 −0.158 0.352 0.568 −0.032 0.396 0.005 0.028 0.284 0.088

0.275 −0.052 0.125 0.116 −0.213 −0.233 0.113 0.288 −0.272 0.834 −0.192 0.196 0.040

1.426 2.256a 1.084 1.326 0.959 3.390b 6.217c 3.080b 2.957b 2.313a 0.205

Function

Wilk’s lambda

χ2

d.f.

Significance

1–3 2–3 3

0.772 0.884 0.962

61.866 29.522 9.348

33 20 9

0.002c 0.078a 0.406

Predictor variables

Group means and proportions Local

Regional

UK

Overseas

Firm Age Size (FTEs) R&D expenditure (%) Percentage of technicians Percentage of QSEs Percentage of professionals/managers Export (as % of T/O) Novel products (%) Incremental products (%) Novel processes (%) Incremental processes (%)

42.07 42.40 11.9 0.0271 0.0076 0.3226 0.070 33.3 42.9 14.3 38.1

26.95 62.44 14.0 0.0485 0.0281 0.1918 0.046 47.4 21.1 12.3 40.4

36.73 52.60 15.0 0.0443 0.0142 0.2071 0.054 48.8 22.5 28.8 35.0

41.07 80.20 23.2 0.0664 0.0253 0.1738 0.174 62.3 20.3 23.2 40.6

N

42

57

80

69

NB: 300 firms had no innovation links. a Significant at 10% level. b Significant at 5% level. c Significant at 1% level.

appear more likely to have collaborative partners located at a higher (i.e. more geographically distant) spatial level than do incremental innovators (product only). Whilst there is some empirical precedent for this result (Kaufmann and Todtling, 2000; Koschatzky, 2000), it is, in general, somewhat counter-intuitive and, indeed, runs counter to accepted wisdom. The ‘milieux innovators’ approach (see for example, Maillat, 1991, 1995; Maillat and Lecoq, 1992) for

example, suggests that the local environment is of little importance for firms developing incremental innovations, since resources required for this kind of innovation are easily found within the firm. By contrast, the local environment is key to the successful development of novel innovations. The more novel the innovation the more important spatial proximity becomes. The data presented here fails to support this thesis. Moreover, R&D expenditure, as an alternative

766

M.S. Freel / Research Policy 32 (2003) 751–770

Fig. 4. Function at group centroids.

proxy for novelty of innovation, does not appear to effect the spatial distribution of innovation-related linkages.

6. Concluding remarks Employing a sample of 597 small manufacturing firms, this paper sought to explore two central issues: the degree of association between external linkages and firm-level innovation performance, on the one hand, and the extent to which such linkages were clustered spatially, on the other. To this end, the empirical analysis presented here indicates a number of interesting results. In the first instance, with regards to the former issue, one notes a highly-varied pattern of association across Pavitt’s sectoral taxonomy and across innovation type. The corollary, which may be drawn from this, is that ‘networks’ (in the sense used here) are not homogeneous or homogeneously good. The influence of various types of innovation-related cooperative networking is likely to differ depending, inter alia, upon the availability of internal competencies, the type of user, the balance between product and process innovation and, relatedly, the underlying technology trajectory (including issues of appropriability and cumulativeness). Thus, pointing to considerably

more complexity than one might infer from policy pronouncements, such as that noted in the recent UK Competitiveness White Paper (DTI, 1998) (quoted earlier), which may reasonably be paraphrased as: ‘Networks are good, more networks are better’. Moreover, it is imperative that one does not neglect the relationship between innovativeness (in both products and processes) and the presence of extensive internal resources. Here again, the data from the current study indicate considerable sectoral variation and variation by innovation type—largely in line with expectations. However, looking simplistically at the all-sample models, R&D expenditure and technical employment were strongly positively associated with successful novel product and process introduction at the firm level. There can be little doubt that internal resources frequently act as complement to, and more occasionally a substitute for, extramurally sourced technology. For science-based firms, where no statistical associations relating to internal resources were observed, the models, themselves, were found to be poor predictors of ‘innovativeness’—suggesting some source of firm heterogeneity not captured by the regressors, perhaps in the form of internal transformation activities. Moreover, and placing these specific remarks to one side, the sample data suggests a further, more fundamental issue, which should give pause for thought. Though, clearly, the network approach to innovation—

M.S. Freel / Research Policy 32 (2003) 751–770

‘. . . that innovating firms are not islands of planned co-ordination in a sea of market relations’ (Oerlemans et al., 1998, p. 307)—is not without substance, one must include an important codicil; a great many of the current sample firms appear to have successfully innovated without requiring ad hoc collaboration or articulation with any other organisations. Accordingly, indeed crucially, it is likely ‘. . . that internal strategies relying on the firms’ own capacities are significantly more important than strategies involving external partners’ (Kaufmann and Todtling, 2000, p. 33).4 Remarkably, 53% of science-based firms (i.e. those for whom external knowledge and external interactive learning is considered particularly relevant), within the current sample, record no external collaboration during the survey period. It seems external collaboration is, unequivocally, neither a necessary nor less a sufficient condition for successful innovation. To reiterate, inferences from the current research coincides with the observations of, inter alia, Arndt and Sternberg (2000) and Oerlemans et al. (1998); namely, that the economic network approach overstates the role of external factors in the innovation process. ‘[I]n most industries the lions share of innovation effort is made by firms themselves’ (Nelson, 2000, p. 13) and occurs within firms themselves. One is tempted to accept Oerlemans et al.’s (1998) contention that, in most instances, ‘. . . innovation is primarily a process built on internal capabilities’ (p. 308), which may, more occasionally, be complemented by external agency. Acceptance of this postulate has, at least, one significant implication. That is, since we know innovative activity to be largely incremental, building upon current capabilities and prior experiences and involving some tacit dimension (in the sense that it is facilitated by ‘learning by doing’ or ‘learning by using’), and that tacit knowledge, in its turn, is essentially person embodied, then ‘. . . any firm level strategy for the development of knowledge must therefore be an employment strategy’ (Smith, 2000, p. 89). As Nelson (2000) noted from a series of case studies, the presence of education and training systems which ensured a flow of individuals with the requisite knowledge and skills, was a distinguishing feature of those countries that were able to beget and sustain competitive and 4 The absence of more specific internal resources may also account for the low ‘explanatory’ power of the models presented.

767

innovative firms. However, an important caveat would be to note that ‘. . . while strength in “high-tech” depends upon the availability of university trained people, industry more generally requires a supply of literate, numerically competent, people in a wide range of functions outside R&D’ (Nelson, 2000, p. 19). Following this, one of the persistent criticisms of UK competition policy, and competitiveness, relates to the paucity of technically skilled individuals at intermediate levels (see Oughton, 1997). In the current study, and with the exception of product innovation in production-intensive firms, it is the employment of technicians, rather than QSEs, which is generally positively associated with ‘innovativeness’. Finally, with respect to the spatial distribution of linkages, there is little evidence that these are clustered geographically (see also Fig. 5). This, however, does not imply that the nature of the innovation process does not impact upon the spatial distribution of innovation partners. Rather, whilst the data provides further supporting evidence regarding the role played by firm size and product markets (Arndt and Sternberg, 2000; Kaufmann and Todtling, 2000), it also indicates a significant relationship between scope of innovation and the reach of linkages. Again, this, in itself, is not remarkable. The general view holds that the local environment, and the relative propinquity of innovation partners, is likely to be more important for novel innovators than for incremental innovators (Baptista and Swann, 1998; Nooteboom, 1999). Yet, the current sample points to the opposite relationship. Novel innovators (i.e. those introducing products or processes new to the industry) are marked by the greater geographical reach of their innovation networks, whilst incremental product innovators appear to be more locally embedded. This may suggest that ‘. . . the probability that local ties can offer all complementary resources is low’ (Oerlemans et al., 2001, p. 4). However, it also raises concerns over the appropriateness of cluster-driven network formation policies. The UK’s experience of such policies, most clearly manifest in the science-park phenomenon (cf. Garnsey and Cannon-Brookes, 1993; Westhead and Storey, 1995; Westhead, 1997) has met with limited success. Finally, it is worth noting that, while Cooke and Morgan (1994) lament the conventional economics view of firms ‘. . . contextualised in terms of industries, sectors and markets’ (p. 25)—considered

768

M.S. Freel / Research Policy 32 (2003) 751–770

Fig. 5. Spatial distribution of innovation-related linkages.

‘spatially insensitive’—and suggest that the adoption of a regional perspective would be more appropriate (at least in the context of industrial innovation), the data from the current study serves, largely, to legitimise the former in preference to the latter. That is, vertical industry, sector and market links clearly dominate (recall Fig. 2). Even in the great exemplars of agglomeration-based regional development (such as Silicon Valley and the ‘Third Italy’), clusters are largely defined sectorally or, indeed, by product markets and not by some abstruse notion of ‘high-tech’, ‘bio-tech’ or some such thing. It is essential that, in developing strategies to facilitate the creation of networks or strengthen a given innovation system, policy pays sufficient heed to this premise. In closing it is important to note a limitation: the data employed here is cross-sectional (incorporating some limited retrospective reporting on the part of respondents). Accordingly, any causal inferences drawn must be treated with care.

Acknowledgements The data presented here was collected as a part of a survey of ‘Enterprise in Northern Britain’. Accordingly, I am grateful to the individuals and institutions involved in this project. In addition, I am grateful to the participants at the Fifth EUNIP conference in Vienna

at which an earlier draft of this paper was presented. Finally, I am grateful for the constructive comments of two anonymous reviewers. Nonetheless, any errors in analysis and interpretation remain my own.

References Acs, Z., Audretsch, D., 1987. Innovation, market structure and firm size. The Review of Economics and Statistics 69, 567–574. Acs, Z., Audretsch, D., 1988. Innovation in large and small firms: an empirical analysis. The American Economic Review 78, 678–690. Acs, Z., FitzRoy, F., Smith, I., 1994. High-technology employment and university R&D spillovers: evidence from US cities. Occasional paper no. 50. CIBER, University of Maryland. Andréosso-O’Callaghan, B., 2000. Territory, research and technology linkages—is the Shannon region a propitious local system of innovation. Entrepreneurship and Regional Development 12, 69–87. Archibugi, D., Howells, J., Michie, J., 1999. Innovation systems and policy in a global economy. In: Archibugi, D., Howells, J., Michie, J. (Eds.), Innovation Policy in a Global Economy. CUP, Cambridge, pp. 1–18. Arndt, O., Sternberg, R., 2000. Do manufacturing firms profit from intra-regional innovation linkages? An empirical-based answer. European Planning Studies 8, 465–485. Arrow, K., 1962. The economic implications of learning by doing. Review of Economic Studies 29, 155–173. Audretsch, D., 1995. Innovation and Industry Evolution. MIT Press, Cambridge, MA. Baptista, R., Swann, P., 1998. Do firms in clusters innovate more? Research Policy 27, 525–540.

M.S. Freel / Research Policy 32 (2003) 751–770 Becattini, G., 1978. The development of light industry in Tuscany. Economic Notes 2–3, 53–78. Becattini, G., 1990. The Marshallian industrial district as a socioeconomic notion. In: Pyke, F., Becattini, G., Sengenberger, W. (Eds.), Industrial Districts and Inter-Firm Co-operation in Italy. International Institute for Labour Studies, Geneva. Bianchi, G., 1990. Innovating in the local system of small–medium-sized enterprises. Entrepreneurship and Regional Development 2, 57–69. Bianchi, G., 1998. Requiem for the Third Italy? Rise and fall of a too successful concept. Entrepreneurship and Regional Development 10, 93–116. Brown, B., Butler, J., 1993. Networks and entrepreneurial development: the shadow of borders. Entrepreneurship and Regional Development 5, 101–116. Brusco, S., 1982. The Emilian model: productive decentralisation and social integration. Cambridge Journal of Economics 6, 63– 82. Cohen, W., Levinthal, D., 1989. Innovation and learning: the two faces of R&D. The Economic Journal 99, 569–596. Cohen, W., Levinthal, D., 1990. Absorptive capacity: a new perspective on learning and innovation. Administrative Science Quarterly 35, 128–152. Cooke P., Morgan, K., 1994. The creative milieu: a regional perspective on innovation. In: Dodgson, M., Rothwell, R. (Eds.), The Handbook of Industrial Innovation. Edward Elgar, Cheltenham, pp. 25–32. Cooke, P., Morgan, K., 1998. The Associational Economy: Firms, Regions and Innovation. Oxford University Press, Oxford. Cosh, A., Hughes, A., Wood E., 1998. Innovation Surveys and Very Small Enterprises, WP 89. Centre for Business Research, University of Cambridge. de la Mothe, J., Paquet, G., 1998. National innovation systems, ‘real economies’ and instituted process. Small Business Economics 11, 101–111. Dosi, G., 1988. Sources, procedures and microeconomic effects of innovation. Journal of Economic Literature 26, 1120–1171. Dosi, G., 1999. Some notes on national systems of innovation and production, and their implications for economic analysis. In: Archibugi, D., Howells, J., Michie, J. (Eds.), Innovation Policy in a Global Economy. CUP, Cambridge, pp. 35–48. DTI (Department of Trade and Industry), 1998. Our Competitive Future: Building the Knowledge Economy, White Paper, Cm 4176. HMSO, London. DTI (Department of Trade and Industry), 2001a. Business Clusters in the UK —A First Assessment. DTI, London. DTI (Department of Trade and Industry), 2001b. The Department of Trade and Industry Science and Innovation Strategy 2001. DTI, London. Etzkowitz, H., Leydesdorff, L., 2000. The dynamics of innovation: from National Systems and “Mode 2” to a Triple Helix of university–industry–government relations. Research Policy 29, 109–124. Feldman, M., 1994. Knowledge complementarity and innovation. Small Business Economics 6, 363–372. Florida, R., 1995. Towards the learning region. Futures 27, 527– 536.

769

Freel, M., 2000. External linkages and product innovation in small manufacturing firms. Entrepreneurship and Regional Development 12, 245–266. Freel, M., 2002. On regional innovation systems: illustrations from the West Midlands. Environment and Planning C: Government and Policy, in press. Freeman, C., Soete, L., 1997. The Economics of Industrial Innovation, 3rd Edition. Pinter, London. Gardiner, P., Rothwell, R., 1985. Invention, innovation, re-innovation and the role of the user: a case study of British hovercraft development. Technovation 3, 167–186. Garnsey, E., Cannon-Brookes, A., 1993. The ‘Cambridge Phenomenon’ revisited: aggregate change among Cambridge high-technology companies since 1985. Entrepreneurship and Regional Development 5, 179–207. Geroski, P., 1990. Innovation, technological opportunity, and market structure. Oxford Economic Papers 42, 586–602. Geroski, P., Machin, S., 1992. Do innovating firms outperform non-innovators? Business Strategy Review, Summer, 79–90. Geroski, P., Machin, S., 1993. Innovation, profitability and growth over the business cycle. Empirica 20, 35–50. Granovetter, M., 1982. The strength of weak ties: a network theory revisited. In: Marsden, P., Lin, N. (Eds.), Social Structure and Network Analysis. Sage, Beverly Hills, CA, pp. 105–130. Håkansson, H., 1987. Product development in networks. In: Håkansson, H. (Ed.), Industrial Technological Development: A Network Approach. Croom Helm, London, pp. 84–127. Hansen, N., 1990. Innovative regional milieux, small firms and regional development: evidence from Mediterranean France. Annals of Regional Science 24, 107–123. Hoffman, K., Milady, P., Bessant, J., Perren, L., 1998. Small firms, R&D, technology and innovation in the UK: a literature review. Technovation 18, 39–55. Howells, J., 1995. Tacit Knowledge and Technology Transfer, WP 16. Centre for Business Research, University of Cambridge. Howells, J., 1999. Regional systems of innovation? In: Archibugi, D., Howells, J., Michie, J. (Eds.), Innovation Policy in a Global Economy. CUP, Cambridge, pp. 67–93. Johnson, B., Lundvall, B., 1993. Catching-up and institutional learning under post-socialism. In: Hausner, J., Jessop, B., Neilsen, K. (Eds.), Institutional Frameworks of Market Economies. Avebury, Aldershot, pp. 68–86. Karlsson, C., Olsson, O., 1998. Product innovation in small and large enterprises. Small Business Economics 10, 31–46. Kaufmann, A., Todtling, F., 2000. Systems of innovation in traditional industrial regions: the case of Styria in a comparative perspective. Regional Studies 34, 29–40. Kirat, T., Lung, Y., 1999. Innovation and proximity: territories as loci of collective learning. European Urban and Regional Studies 6, 27–38. Kline, S., Rosenberg, N., 1986. An overview of innovation. In: Landua, R., Rosenberg, N. (Eds.), The Positive Sum Strategy: Harnessing Technology for Economic Growth. National Academic Press, Washington, DC. Koschatzky, K., 2000. The Regionalisation of Innovation Policy in Germany—Theoretical Foundations and Recent Experience, WP R1/2000. Fraunhofer ISIR.

770

M.S. Freel / Research Policy 32 (2003) 751–770

Lawson, C., 1999. Towards a competence theory of the region. Cambridge Journal of Economics 23, 151–166. Love, B., Roper, S., 1999. The determinants of innovation: R&D, technology transfer and networking effects. Review of Industrial Organisation 15, 43–64. Love, J., Roper, S., 2001. Location and network effects on innovation success: evidence for UK, German and Irish manufacturing plants, German and Irish manufacturing plants. Research Policy 30, 643–662. Lundvall, B., 1995. National Systems of Innovation: Towards a Theory of Innovation and Interactive Learning. Pinter, London (first published in 1992). Maillat, D., 1991. Local dynamism, milieu and innovative enterprises. In: Brotchie, J., Batty, H., Hall, P., Newton, P. (Eds.), Cities of the 21st Century. Longman, London, pp. 265–274. Maillat, D., 1995. Territorial dynamic, innovative milieus and regional policy. Entrepreneurship and Regional Development 7, 157–165. Maillat, D., Lecoq, B., 1992. New technologies and transformation of regional structures in Europe: the role of the milieu. Entrepreneurship and Regional Development 4, pp. 1–20. Nelson, R., 2000. National innovation systems. In: Acs, Z., (Ed.), Regional Innovation, Knowledge and Global Change. Pinter, London, pp. 11–26. Nooteboom, B., 1999. Innovation, learning and industrial organisation. Cambridge Journal of Economics 23, 127–150. Oerlemans, L., Meeus, M., Boekema, F., 1998. Do networks matter for innovation? The usefulness of the economic network approach in analysing innovation. Tijdschrift voor Economische en Sociale Geografie 89, 298–309. Oerlemans, L., Meeus, M., Boekema, F., 2001. Firm clustering and innovation: determinants and effects. Papers in Regional Science 80, 337–356.

Oughton, C., 1997. Competitiveness policy in the 1990s. Economic Journal 107, 1486–1503. Oughton, C., Whittam, G., 1997. Competition and co-operation in the small firm sector. Scottish Journal of Political Economy 44, 1–30. Pavitt, K., 1984. Sectoral patterns of technical change: towards a taxonomy and a theory. Research Policy 13, 343–373. Polanyi, M., 1966. The Tacit Dimension. Routledge and Kegan Paul, London. Rosenberg, N., 1982. Inside the Black Box: Technology and Economics. Cambridge University Press, Cambridge. Rothwell, R., 1991. External networking and innovation in small and medium-sized manufacturing firms. Technovation 11, 93– 112. Shaw, B., 1991. Developing technological innovations within networks. Entrepreneurship and Regional Development 3, 111– 128. Staber, U., 1997. An ecological perspective on entrepreneurship in industrial districts. Entrepreneurship and Regional Development 9, 45–64. Sternberg, R., 2000. Innovation networks and regional development—evidence from the European Regional Innovation Survey (ERIS): theoretical concepts, methodological approach, empirical basis and introduction to the theme issue. European Planning Studies 8, 389–407. Tabachnik, B., Fidell, L., 2001. Using Multivariate Statistics, 4th Edition. Pearson, Needham Heights, MA. von Hippel, E., 1978. Successful industrial products from customer ideas. Journal of Marketing 42, 39–49. Westhead, P., 1997. R&D ‘inputs’ and ‘outputs’ of technologybased firms located on and off Science Parks. R&D Management 27, 45–62. Westhead, P., Storey, D., 1995. Links between higher education institutions and high-technology firms. OMEGA 23, 345–360.